Problem about "pair_style srp"

LAMMPS LAMMPS Mailing List Mirror
1

Dear LAMMPS users,

I’m performing “pair_style SRP” calculation for polymer chain and water system, but my calculation stops without any message. I don’t know why my calculation stops.

My input file is shown below;

2

Hi Hironori,

Which version of LAMMPS do you use? There are some bugs for pair_style srp in the earlier version of LAMMPS. I suggest you use the latest version of LAMMPS to run it again.

Best,

MC

3

Hi Mingchao

I forgot to show version. Thank you for your reply. My LAMMPS version is “LAMMPS (16 Feb 2016)”. This must be the latest version of LAMMPS.

Best,

Hironori Sakai

4

Hi Mingchao
I forgot to show version. Thank you for your reply. My LAMMPS version is
"LAMMPS (16 Feb 2016)". This must be the latest version of LAMMPS.

​the very latest version of LAMMPS is 23 March 2016. you can see this and a
summary of the changes per patch here: ​http://lammps.sandia.gov/bug.html

to address the issue you are reporting. there *has* to be some output or
message. it may not be in the log file. also, some MPI libraries are not
always returning all output to the frontend machine. so please test in
serial execution, if needed with the help of a debugger.
if you still cannot find out the source of the issue, please make a full
input deck available. ideally one that takes the least possible amount of
time until the crash/hang situation occurs.

​axel.​

5

Dear Axel,

Thank you for your reply.
As you proposed, I have tested my calculation in serial execution (lmp_serial). As a result, my calculation was terminated normally. In my computational environment, other calculations, for example DPD or atomic MD were normally calculated in MPI execution. Only SRP calculation seems to be not performed in lmp_mpi.

I made small input file set for representation my problem. This calculation is terminated normally in serial execution, but in MPI execution, calculation stops. I used version “LAMMPS (16 Feb 2016)”. I know this version is stable so I’m using it.

Best,

Hironori Sakai

in.DPD_Naf (882 Bytes)

101010.lmp (173 KB)

6

Dear Axel,

Thank you for your reply.
As you proposed, I have tested my calculation in serial execution
(lmp_serial). As a result, my calculation was terminated normally. In my
computational environment, other calculations, for example DPD or atomic MD
were normally calculated in MPI execution. Only SRP calculation seems to be
not performed in lmp_mpi.

I made small input file set for representation my problem. This
calculation is terminated normally in serial execution, but in MPI
execution, calculation stops. I used version "LAMMPS (16 Feb 2016)". I know
this version is stable so I'm using it.

​i am having no problem running this input in parallel. what is your MPI
library, what kind of machine are you running on and what is the exact
command line?

BTW: for better performance, i would recommend trying the following
changes. replace:

neighbor 2.0 bin
neigh_modify every 1 delay 0 check yes one 240000 page 2800000

with:

comm_modify cutoff 3.0
neighbor 1.0 bin
neigh_modify every 1 delay 0 check yes

​this will extend the communication cutoff as required without having to
increase the neighbor lists.​

​axel.​

7

Dear Axel,

Thank you for your comments.
About computational setting, I will ask our system administrator.

Best,
Hironori Sakai